In order to apply gene-order rearrangement algorithms to the comparison of genome sequences, Pevzner and Tesler bypass gene finding and ortholog identification and use the order of homologous blocks of unannotated sequence as input. The method excludes blocks shorter than a threshold length. Here we investigate possible biases introduced by eliminating short blocks, focusing on the notion of breakpoint reuse introduced by these authors. Analytic and simulation methods show that reuse is very sensitive to the proportion of blocks excluded. As is pertinent to the comparison of mammalian genomes, this exclusion risks randomizing the comparison partially or entirely.